Machine learning vs. traditional methods for predicting postoperative cardiac complications after non‐cardiac surgery: a systematic review and Bayesian network meta‐analysis

The study compared the ability of machine learning and traditional methods to predict cardiac complications after surgical procedures. The analysis included 13 studies with 927,113 patients and 54 predictive models. Machine learning models generally outperformed traditional scoring systems, with automated machine learning achieving the highest ranking (SUCRA 96.6) with a mean difference of 0.28. Gradient boosting models also demonstrated superior performance with a difference of 0.20 compared to the standard Revised Cardiac Risk Index. The main limitation was the absence of external validation of machine learning models and low quality of most included studies. The authors conclude that prospective multicentre studies are needed before clinical implementation.